Paula Rivera:
Welcome to the AI Factor where we explore how artificial intelligence is shaping the way businesses work. I’m your host, Paula Rivera, and today we’re diving into one of the most exciting frontiers in the AI space, agentic AI. There’s a lot of buzz around AI lately, but also a share of confusion. What does agentic AI even mean? How is it different from the bots and automation we already know? And more importantly, how are forward-thinking companies putting it to work?
To help demystify it all, I’m joined by two incredible minds from IntelePeer’s AI application engineering team, Cora Brito and Erica Rusch. They’ve been at the forefront of building real-world agentic systems, and today they’re here to give us an inside look.
Ladies, welcome.

Cora Brito:
Thank you for having us.

Erica Rusch:
Glad to be here.

Paula Rivera:
Wonderful. I’m so excited to be talking to you guys. It’s always a fun conversation.
Before we jump into the heart of the discussion, I just wanted to get a little bit of a fun opener going on here. Taylor Swift recently won a lawsuit that basically made it illegal for people to use AI images specifically for revenge purposes. Think revenge porn. Great, amazing lawsuit to win for anybody, especially us ladies out there.
But likewise, OpenAI recently brought back the Sky voice after Scarlett Johansson controversy, saying it wasn’t a copy. They had paused Sky on ChatGPT after Scarlett claimed it sounded too much like her, raising questions about consent, AI voice ethics, and how far is too far in a synthetic personality creation.
Cora, voice is becoming one of the most human-like elements of AI, we’ve talked about this in the past, and it’s sparking a lot of debate. When you hear a voice like Sky, do you find it helpful, creepy, or part of the evolution?

Cora Brito:
I think it’s fascinating how voice is like a bridge now between technology and human connection. We’re starting to see more and more cases like this from AI voices and scenarios, but truly it’s a consent and authenticity matter where is it unique to that person? And I almost think it’s almost like a signature. So would you be okay having somebody else sign your documents? So, I think the voice is definitely very personal.
And so is it creepy? Maybe, but is it crossing the line? You kind of have to be careful with that because it is an ethical line, but yeah, we’re going to see a lot more of these cases.

Paula Rivera:
Yeah, I like the fact that you bring up would you let someone else sign your documents? Your voice could almost be considered part of your fingerprint or your overall person print, so very valid point. Thank you so much.
Erica, if you could give your own agentic AI assistant any voice, real or fictional, who would it be and why?

Erica Rusch:
Oh, wow. What a cool question. I think maybe Ellen Pompeo from Grey’s Anatomy. She has the most comforting voice, matter of fact, clear, articulate, so you always know what she’s saying, but it doesn’t sound so robotic.

Paula Rivera:
I love that, and you picked one of my favorite shows, so I appreciate that.
Let’s start with some myth-busting. As AI is everywhere, but there’s still a lot of confusion about what it is and isn’t, let’s break things down a little bit. Cora, what are some of the biggest misconceptions you hear about AI, especially in the enterprise world?

Cora Brito:
Oh, there’s so many. One that comes to mind is where people will think that all AI is the same. It’s like using the word AI is covering all descriptions, and I think that it’s not one technology. It’s a great technology that has multiple uses, but it’s not one-shoe-fits-all.
A good way to understand that this is just a wrong interpretation, thinking of getting from Point A to Point B, and what would you use to get from Point A to Point B would matter on what you need to do. Would you get a bike or would you get a rocket ship? So that’s how, I guess, different AI could be used in different scenarios.
Another misconception, I think, is it’s a magic employee. If they just get an AI-automated feature or something that is being done, then they believe that that’s the “set it and forget it,” like you don’t have to go back again, and that’s a big mistake. The AI in many enterprises will be like a intern, very, very smart, but you still need to provide it all the training and monitoring that goes behind it. So it’s faster than a human, but it definitely needs some instructions and a manager and oversight over it.

Paula Rivera:
So I love the fact that you bring that up because the other day I was doing an internal email, and I was like, “Oh, let me just have AI do this because it would be quicker than doing it myself.” So, they did it. At first glance, it looked well and I was like, “Okay, this is good.” And then I went back and re-looked at it, and I’m like, “Oh my goodness, this is not what I asked it to do at all.” I’m like, “Where did they get this information?” I basically had to throw that out and I just wound up doing it myself. To get what I needed was just taking way too much time. That being said, more often than not, it does reduce time for me, but you just have to be super careful with it. I really appreciate that, Cora.
Erica, how would you explain the difference between traditional automation in today’s advanced AI to a business leader or maybe someone like myself?

Erica Rusch:
So traditional automation, one of the easiest examples is an assembly line. Another example of automation would be the traditional IVR or even a Google Dialogflow hard-coded chatbot. What makes advanced AI in business different and what makes it stand out and why people want it is because it’s adaptable to the real-life nuances of a live interaction. It’s not a hard-coded intent. We don’t have to tell it exactly what the customer is going to say. We don’t have to program every correct response. It gets to adapt to the conversation in real time, and that’s what makes it so powerful.

Paula Rivera:
Oh, I love that and I appreciate it, and this is where I throw the questions out the window. As you were speaking, you made me think of, you would think of a Siri or an Alexa as, I mean, technically it’s AI, but it’s primitive AI and I would say it’s more like automation than it is today’s generative AI.
How far away do you think those two kind of uses of AI will adopt the new technology and become better and more advanced?

Erica Rusch:
You’re asking how long I think Siri or any of the other virtual assistants will start to use AI to be more helpful?

Paula Rivera:
Well, more advanced AI because technically it’s always been AI, just primitive AI.

Erica Rusch:
I don’t think we’re far away from that at all. I think Gemini just released an updated version of a virtual assistant. So did Samsung, I think, just release a more updated version of a virtual assistant. I think that we’re going to start seeing that happen rapidly, not months or years, days and weeks.

Paula Rivera:
Yeah, so thank you. I appreciate that. And I am in the camp of, if the consumers start using it, businesses will be forced to use it. So you need something like a Siri or an Alexa to really harness the most recent generative AI and AI technologies, and you need the consumer clamoring for it in order for businesses to truly embrace it. Now, I could be a little bit wrong, but I do think that speeds progress along a little bit. I don’t know if either of you have thoughts on that.

Erica Rusch:
Oh, absolutely. The second the consumer wants something and they would be willing to pay for it or use it in a way that would generate resource for the company, whether that’s data or money, it hits the market quite quickly.

Paula Rivera:
Yep, I appreciate that. Super helpful. I appreciate clearing the air. Let’s get to the main event, agentic AI.
Erica, can you define it for us? What is agentic AI in simple terms?

Erica Rusch:
Yeah. So what makes Agentic AI different than traditional prompting is when you write a traditional prompt for an LLM, you’re giving it like a job description, “This is what I want you to do, the way I want you to do it, with these guidelines.” What you’re doing with agentic AI is you’re giving not only the job description but the next five years’ of experience along with it because you get to break up the instructions into bite-sized pieces so that you can provide relevant information, like ASR configurations or a specific variable, at that place so that the agent can do a really good job with the task that you’re giving it on hand, and then you’re telling it, “These are all of the different ways that the conversation can go from here, and here’s how I want you to handle those.”
So it’s taking the power of building out a really beautiful conversation design flowchart and pairing it with the adaptability of generative AI in a conversation and giving it this really robust set of tools so that it can handle conversations in a very directable, maintainable, flexible, and scalable way. That’s what makes it so powerful for enterprise-level solutions.

Paula Rivera:
Wow, that was beautifully said.

Erica Rusch:
Thank you.

Paula Rivera:
Well, thank you. You guys are making my job so easy.
Cora, what capabilities do agentic AI unlock that previous models didn’t?

Cora Brito:
Just to continue on with what Erica’s saying with the agentic AI and what makes a difference from the previous models is what she was addressing, and I think the difference for an enterprise customer, like what was the biggest difference? And I want to compare this to maybe giving hands to Alexa to actually grab and do things for you, and those are the tools that she’s talking about.
So, what does that mean for an enterprise customer or what does that mean for a business owner? That means that not just being able to reformat or summarize your emails or just write you a really nice cover letter or a work letter, those things we know it can handle, but now we’re actually telling it to do things when it gets that information, and that’s the agentic side of it. So it’s almost like where humans are constantly juggling multiple tasks, that’s where the AI agentic could really help out automate some of that.

Paula Rivera:
Nice, okay. So it’s one thing to understand the concept, and you guys have laid it out so nicely. Let’s talk about how agentic AI is implemented and what the differences between playing with AI, probably what I do on a daily basis, and/or developing a solution that will have a real impact on the business, what you guys are doing on a daily basis.
Erica, you said that a agentic AI is like an AI prompt that’s structured like a map. Explain this to me.

Erica Rusch:
Yeah. So when we’re talking about prompting in this way, we’re not just giving the agent a list of things to do or a description of what to expect. We’re also telling it where to go when it gets a certain set of information. And sometimes that looks like from a point where I’m collecting information set A, I have three potential paths, and each of those paths now have their own directions for what to do if my customer and my caller ends up on that path.
When we’re saying we’re giving it a map, we really are. We’re giving it a map of the different ways that we’re expecting a conversation to go, and we’re also telling it what to do if it doesn’t go that way. And then we’re leaving it up to the agent, the LLM, to be smart enough to pick that path and use them all in a logical, adaptable, nuanced way.

Paula Rivera:
So what if the LLM isn’t intelligent enough?

Erica Rusch:
Then we haven’t it enough context or our map is unclear. All of these LLMs that are on the market today are absolutely powerful enough to make these decisions in a way that makes sense and is repeatable not only five or 10 times for you or I using it, but on an enterprise level, hundreds of thousands of times over and over again, if we’re sending in appropriate context and appropriate directions.

Paula Rivera:
Okay, I appreciate that. Tell me a little bit about prompting, and does it help to be uber specific or can you be a little bit on the generic side? What’s the best way to go about prompting so that you get the best result that you’re looking for?

Erica Rusch:
It depends on your goal.

Paula Rivera:
Ah.

Erica Rusch:
Prompting is absolutely, I’ve heard people say more of an art than a science, but it’s an art like persuasive writing is an art because it’s something that you do very intentionally. If I want a very specific result, I give it a very specific instruction, but if I want it to be more adaptable to what the user is going to say, say like, “Can you explain this part of your legal disclaimer to me?” then I give it a more general, open-ended set of instructions.

Paula Rivera:
Okay. And it’s funny that you say that. I started thinking about my husband because I listened, whatever AI tool he’s using, I listened to him give it a prompt, and he’s like, “Write a letter for me.” And I’m like, “Oh, honey, you need to be a little bit more specific than that.”

Erica Rusch:
Yeah, exactly.

Paula Rivera:
And I think actually it may have been the class you taught, Erica, where you tell it, “This is what you are. You’re a communications professional,” or, “you’re a music teacher,” which is what my husband is, and, “this is what you’re trying to do.” So you almost set the stage for the AI system to understand what you’re looking from it.

Erica Rusch:
Exactly. Yeah, and that’s what I meant by saying, “It’s an art like persuasive writing is an art,” because there’s a strategy behind it that you’re very intentional in giving it that profile, not because the AI really cares about who it is or what it’s doing, but because as the person writing the prompt, it’s a very efficient way to give it context and limit tokens and put it on the right path.

Paula Rivera:
Nice. I like your paths and your maps.
Cora, are there any particular industries or business functions that are more ripe for agentic AI than others?

Cora Brito:
Oh, yes, of course. There’s many, many, many, many examples. Some of the top industries that do very, very well with some awesome technology like this is definitely going to be in the customer service area. These are going to be places where you have to understand that the customer agents are doing so much, so many tasks, and some of those tasks could be automated, and that’s where agentic AI really helps out.
In the financial, this is huge. FinTech for fraud detection, loan processing, investment decisions, portfolio management, risk management, making recommendations based on patterns, and the FinTech is a great industry for this.
Another great industry where agentic AI is definitely going to be used in is going to be the healthcare administration side, so insurance verification, patient follow-up, scheduling. That’s one of the projects that we’re currently working right now where a scheduling app is doing lots of automation for dentists’ office, for medical offices, for clinics, and we’re finding that it helps schedule appointments a lot faster, therefore bringing more money for the healthcare industry.

Paula Rivera:
Yeah, I’m a big fan of the reminders because I have my personal calendar and then I have my work calendar and I’m not consistent on where I put stuff. So whenever I get a automated reminder, I am so appreciative of it because I like to be courteous and not-not show up for stuff or not show up for stuff, so that’s super interesting.
Is there any kind of segment of healthcare where you’re seeing this more or is it just across the board?

Cora Brito:
We’ve seen AI really in healthcare for a long time now, but the agentic side of it is where now we’re giving it some tools to be able to do those tasks that would need a human interaction. So it’s doing a lot of connecting of multiple systems and automating a lot of that task that had to do where an agent would take in, look at the schedule, talk to the doctor, see if there’s any availability. All that’s pretty much automatically done, and then whatever tasks you want it to do in the healthcare, so maybe update or send a notification somewhere else.
And all of those points or those tasks that it’s doing, we have to provide it some information. So there’s still some development obviously that needs to happen, but once that’s said and done, it will help small business and healthcare and many businesses automate a lot of tasks.

Paula Rivera:
And I’m inclined to say medical practices, whether it’s your doctor’s office or your dentist’s office, they probably appreciate that. From what I’m reading, there is definitely, I’m going to call it a labor shortage. I don’t know if it’s a labor shortage or lack of talent, for lack of a better word, but these businesses are having a hard time filling front-of-office positions, getting those folks who they’re welcoming people that are coming into the doctor’s office, they’re answering phones and they’re shuffling schedules and they’re doing multiple things.
So that, I can totally see where agentic AI would just take such a burden off of that front-office staff, off of the medical practice in general, and really allow the people who are working there more time to focus on us, the patient. So, I like that. I like what we’re doing.
Obviously, a lot of use cases, a lot of opportunities for agentic AI, super powerful. I mean, I don’t think you can go a day without hearing or reading something about agentic AI. What makes it better than the status quo? Cora, what are the biggest benefits for businesses? Is it cost, scalability, I don’t know, customer experience?

Cora Brito:
All three.

Paula Rivera:
Okay!

Cora Brito:
No, it’s true though when I say, “All three” because costs, right? So maybe it is a heavy upfront cost for a business to implement some of these solutions into their businesses, but over time, their ROI definitely compounds and it’ll be saving them money and actually making the service growing internally, but then also the capacity in what it can do.
As far as scalability, obviously with AI, we always take an approach where you start at scale and then you add as you go. So definitely these are scalable solutions that you can add as you go and as your needs changes. So for a business, cost savings, scalability, and overall customer experience, definitely big benefits.

Paula Rivera:
Nice, nice. Erica, Cora just gave us so many examples of the benefits for businesses, and there was a lot of them. Can companies see value? How quickly can companies see value once they implement an agentic AI system?

Erica Rusch:
Oh, so quickly, and what makes building an agentic AI such a big benefit and why everyone is curious about it is that it makes solutions extremely scalable and it makes them customizable to certain industries. So the time to value, so the time between a contract being signed and when you can start billing for usage, is reduced significantly because you can set up a system for healthcare or for finance or for ordering in a more retail or restaurant-type space. And for the most part, the things that need to be automated from company to company in those spaces can be handled by agentic AI because of its own adaptability, the adaptability of using an LLM because you don’t have to hard code every single turn of a conversation.

Paula Rivera:
Interesting. Now, I’m inclined to say though, it probably is a quicker process for those entities that have their houses in order.

Erica Rusch:
Absolutely.

Paula Rivera:
So if you have your processes in place, let’s say everything’s a little bit regimented, things would go much quickly than if you’re a less organized company. Is that a fair statement?

Erica Rusch:
Oh, absolutely.

Paula Rivera:
Okay. What’s the feedback been from our customers or from people you’ve talked to in the industry using these solutions?

Erica Rusch:
So far, our newest customers that are using the agentic AI model are impressed by how quickly we can deliver features and how well the agent handles unexpected or nuanced conversations.

Paula Rivera:
Nice. So, positive feedback?

Cora Brito:
Yeah.

Erica Rusch:
Yeah, so far, so good. We’re thrilled.

Paula Rivera:
Well-

Cora Brito:
Super positive. If I can just add to that, I think that the way that Erica specifically, especially she’s very focused into the agent exactly giving it some good prompts to stay at task. When you’re at an enterprise level, it’s so important to stay at task so somebody’s not trying to call into the agent and have it talk about food when it’s a healthcare or any other topic. So keeping the agent at task is super important, and that’s something that Erica does very, very good. So customers seem to be very happy with this so that, at an enterprise level, they can use it and it’s very efficient.
One other thing I’ll add, as far as time, we’ve seen some of our customers see a return or an improvement within a week where we were automating a lot of those calls that they get, even with a very simple AI agent that handles some questions and does simple tasks, we’ve seen sometimes up to 20, 30% improvement within a month of being able to handle those calls on the agent side and not so much on a human interaction required.

Paula Rivera:
Nice. We’re delivering some real value, which I’m sure our customers appreciate, and anybody out there in the work world would appreciate, so that’s super great to hear.
The road ahead. We’ve been speaking for a while now. Lots of great examples, ladies. Let’s look into the future. Erica, where do you see agentic AI heading in the next 12 to 24 months?

Erica Rusch:
Hmm. I think that we’re going to see agentic AI becoming more and more adaptable at every part of a conversation because we’ll be able to give it more and more instruction around how to handle that voice interaction, and what information you’re allowed to use at this stage, and what additional paths that you can take. It will be more and more and more adaptable the more people use it because we get more use cases that we can level against, and we get to take our industry experience in building automation and apply that to this, especially in regards to ASR and how to use agentic AI with an ASR appropriately and efficiently and reliably.
The other thing I see changing very quickly, becoming more and more configurable. So when we spin up a new client and they want a certain word change, or they want a certain welcome message, or they want different customizations that make the agent feel like their employee, we’ll have more and more ability to refine those things for that customer.

Paula Rivera:
Cora, are there new capabilities or trends on the horizon you’re excited about?

Cora Brito:
Yes, it’s so exciting, especially, well, now that we know that one agent is super smart, and I’m air-quoting, you can’t see those, but yeah, now that we know these AI agents are super smart, I feel like there’s already a trend happening where they’re collaborating intelligence. So one smart agent is talking to another really, really smart agent, and then another smart agent, and so each of those agents are doing many, many tasks or only one specific topic, and so it makes them an expert at that specific task that they have.
And so we’re going to probably, this is where I feel like it’s going, it’s being used a lot in social media also, where you can have agent that is a project manager that overlooks all the other agents’ task. You’re basically getting an agent that looks after an agent that creates the content. The other one maybe edits the videos. And so all of these agents working together and then having one agent be the project manager, so it’s like agents managing agents, I feel like that’s going to be a big thing.
Also, we’re already seeing it, and Erica mentioned it already, the high use of ChatGPT and DeepSeek and all these great LLMs that are coming out are just basically training it even better. So we’re already starting to see where the agents will reply. Maybe before if you asked it a question or you said something, it’ll be like, “I recommend blah, blah, blah.” But now we’re starting to see it, it will reply with, “I recommend X because of these reasons, and then here are the reference to the reasons why.” So it’s not just providing an answer, but it’s giving some feedback or some background as to how it answered. And I think that’s a behavior because more people used it, right?
So that’s already a trend that’s already happening right now, but yes, lots of cool stuff coming through.

Paula Rivera:
You are a trend identifier. Okay, cracking myself up here.
Listen, we need to wind things down a bit, but before we wind things down, I like to do a rapid fire wrap-up. We’ve been ending the show with a few, quick, rapid-fire questions so that people can get to know you guys a little bit better outside of your job titles and the thoughts you have currently. I have three questions here and I just want you guys to say the first thing that comes to mind. You can expound upon it a little bit if you’d like. If you just want to provide me with a one or two-word answer, I am good with that as well, and I am going to start with Erica.
What AI-powered tool or feature you use almost daily?

Erica Rusch:
In my personal life?

Paula Rivera:
You could do personal.

Erica Rusch:
Oh, sure. Anytime I need to write a grown-up document, do administrative tasks for my whole life, why would I do that from scratch when I can go to Gemini or GPT and tell them to write it for me?

Paula Rivera:
I love it. Would you rather have an AI that does your email or your errands?

Erica Rusch:
Email, hands down.

Paula Rivera:
Okay. And what’s a movie or TV show that got AI right or hilariously wrong?

Erica Rusch:
I actually loved the series on Netflix called Love, Death & Robots. There are a couple episodes on there that are specific to AI. One of them that I remember is a post-apocalyptic short about three robots that are exploring some human ruins and making up all of these stories about what our lives were like, and it’s just a really interesting way to look at what the future of AI could possibly be, but sci-fi and it’s for fun.

Paula Rivera:
Ooh, that does sound fun. I’m going to have to check it out.
All right, Cora, onto you. What AI-powered tool or feature you use almost daily?

Cora Brito:
Definitely has to be my Copilot for emails. It does really good at making sure I sound very, how can I say this, like corp, I guess corporate way so that my thoughts initially are not just like, “What is this?” but more intelligent. So, it makes me sound really good.
Aside from Copilot, I would say definitely ChatGPT for many other things. Just recently, and I think this is the funniest thing, I’ve been looking into Western medicine and so I’d send into ChatGPT a picture of my face and I asked it to provide some health feedback and this thing told me all about gut problems and other things that I may have to get tested now on. So, that was an interesting use.

Paula Rivera:
Ooh. Ooh, that’s intriguing.

Erica Rusch:
Quick disclaimer, GPT is not a doctor.

Cora Brito:
Yeah.

Paula Rivera:
Oh, thank you. Thank you. Legal’s going to be calling me up after this. They’re going to be like, “Paula!”

Erica Rusch:
Beware of hallucinations. Use responsibly.

Cora Brito:
Yes. All of that was given to me at the beginning of it, so it was very clear. But there was just some tests and stuff that I’ve never thought of that now maybe I can just bring them up to my doctor, so that was pretty interesting.

Paula Rivera:
Nice. Okay. Would you rather have an AI that does your email or your errands

Cora Brito:
As you asked this question to Erica, I was thinking, “How can it do my errands?” Agentic, I would have to give it so many tools. It’s just so much. But I would prefer, if I do take the time, I would prefer to give it my errands. Emails, as far as I’m concerned, I can handle those. I’m not trying to [inaudible 00:28:22]-

Paula Rivera:
Well, you have Copilot. You have Copilot handling those.

Cora Brito:
I know. Yeah, something else [inaudible 00:28:27].

Paula Rivera:
Excellent. And what’s a movie or TV show that got it right or hilariously wrong?

Cora Brito:
Okay. A few years ago I saw this series. It was a short series. It’s Year Million, and it’s crazy though. It’s on Prime. So it was about… Just really quick, it started out with an accident that their daughter had, and so the parents were asked a question by this drone that circulates over and that’s the 911. And the drone asked it, “Would you want to get the brain downloaded?”
So that was enough right there to hook me on, and it goes into how we’re going to probably live to three, 400 years because we are in this health train where we just want to get healthier and healthier and healthier. So what will be our life expectancy? And it just makes you think about some crazy stuff. Right now it’s called Year Million because we don’t know when, but if everything keeps going the way it is, it seems like it’s trending to have some of those abilities in AI.

Paula Rivera:
Oh, gosh. I now have two bingeworthy shows that I’m definitely going to check out and I really would not want my parents to know what’s in my brain. Okay.

Erica Rusch:
Oh my gosh. Whoa.

Cora Brito:
Wow, Paula, you just opened a whole thought.

Paula Rivera:
On that note, I started out with revenge porn and I’m ending with what’s on my brain.
So I’d like to thank IntelePeer’s Cora Brito and Erica Rusch for helping us decode agentic AI. If you’re looking to level up your business processes or elevate your customer experience, this episode should give you a strong sense of where the future is headed. If you liked today’s conversation, make sure to subscribe and leave us a review. And if you’re curious about how agentic AI could work for your business, check out intelepeer.ai for more information.
Until next time, thanks for listening to AI Factor.